Box Straight Line Detection Method Based on Two PCA and DBSCAN

With the rapid development of modern logistics technology, intelligent forklifts have become an important part of intelligent logistics. Intelligent forklifts based on computer vision are developing continuously. In order to adapt to the working environment of boxed and stored goods, this paper designs a linear detection algorithm combining computer vision and machine learning, so as to apply it to intelligent forklifts to identify boxed goods. position and contour and perform access work. This paper proposes twice principal component analysis(TPCA), and uses TPCA to remove obvious noise and detect straight lines, and combines DBSCAN clustering algorithm to classify its data and do twice principal component analysis. The work of this paper is mainly to perform straight line detection on the edge contour of the boxed goods in the warehousing environment. First, a model for identifying the contour box of the boxed goods is trained in Faster-RCNN, and then the rough outline of the boxed goods identified by the model is extracted. RoI, and perform straight line detection on the data in the RoI. Experiments show that the recognition accuracy in the manually collected pictures is 89.10%, of which the correct box recognition rate is 97.91%, and the picture accuracy rate in the normal environment is 91.61%, of which the correct box recognition rate is 98.40%.